a16z Launches Ops Engineering Fellowship as AI Workflow Roles Proliferate

Key Takeaways

1. a16z is launching an Ops Engineering Fellowship for ~50 internally deployed engineers and "AI-pilled" operators, with a 6-week in-person program in SF and NYC.
2. Noah Kagan connected 22 data sources to Claude and found a broken discount code that had cost $60K.
3. Arvid Kahl argues testing skills—E2E, integration, unit, mutation, fuzzing, static analysis—are now critical in agentic AI codebases.
4. Acquire and Flippa listings show small profitable AI and e-commerce businesses changing hands at modest multiples.
5. A 37-point enterprise AI transformation checklist warns against starting from "we need AI" and stacking AI on old processes.

1. AI Roles and Enterprise Adoption

1. a16z has launched an Ops Engineering Fellowship for roughly 50 internally deployed engineers and "AI-pilled" operators, running as a 6-week in-person program in San Francisco and New York with a group chat. The role connects internal systems, builds agents, and reworks finance, sales, and support workflows; @levie calls these people "Internally deployed FDEs," while others use "AI Operations" or "Agent Engineer," with no unified title yet. Confirmed Founding Fellows include Claire Vo, Sebastien Goddijn (building Glass at Ramp), Owen Williams (Stripe), Ann Miura-Ko (Floodgate), Brandon Gell (Every COO), TK Kong (co-founder of Ramp Labs), and Matt P (internal AI at Decagon), plus undisclosed members. The account is also publicly asking who owns "Ops Engineering" or similar functions at their company and who should join the cohort or a dinner, with an application link. — via 1 2
2. Noah Kagan says his biggest breakthrough this quarter was connecting 22 data sources to Claude, including Stripe, QuickBooks, Ramp, HubSpot, Help Scout, and Klaviyo, enabling single-conversation queries. When he had Claude audit an email welcome sequence, it found a discount code that had stopped sending, an error that cost $60K before the team fixed it. Other uses include generating daily reports and improvement actions, cross-referencing Google Analytics with Metabase to investigate a sales decline, and auditing an affiliate program for opportunities; most tools have built-in connectors, with custom MCP connectors for the rest. He also says Claude is offering $250 in free credits. — via 1 2
3. A forwarded 37-point enterprise AI transformation checklist argues companies should start from real problems rather than "we need AI," avoid stacking AI on old processes, and not over-engineer when traditional software suffices. It stresses executive ownership, CEO sponsorship, early involvement of legal/finance/IT, change management investment, and dedicated experimentation time for employees, warning that missing an AI center of excellence and knowledge-sharing channels leads to shadow AI. On measurement, it says to establish baseline metrics before adoption, accelerate existing KPIs rather than inventing AI KPIs, avoid premature layoffs and harsh ROI tests, and treat AI transformation as a multi-year migration. Data and governance prerequisites include resolving data permissions and RBAC before rolling out agentic tools company-wide, avoiding single-vendor lock-in, and focusing on the "last mile" of governance, data, observability, context management, and process rather than just the model. It also warns against outsourcing thinking: everyone can write prompts, but the difference is how tools amplify work—good judgment multiplies output, poor judgment wastes tokens—and calls for continuous internal evals measuring cost per successful task. — via 1

2. Developer Skills and Agentic Workflows

1. Arvid Kahl says many developers still believe their main job in the coming years will be writing or reviewing code. In an agentic development environment, he argues that to stay employable, developers should master all layers of testing: end-to-end, integration, unit, mutation, fuzzing, and static analysis. In AI codebases of meaningful size, these have moved from "nice to have" to "absolutely critical." — via 1 2
2. A forwarded video outlines a full path to build and ship a working app from scratch using Claude Code, GitHub, and Vercel, emphasizing writing a plan first as the agent's "GPS," designing before coding, and distinguishing "build vs. buy"—only build what gives you an edge over others. — via 1

3. Acquisitions and Small Business Listings

1. Acquire has listed several small businesses: a health-focused repeat-purchase e-commerce store with $29K TTM revenue asking $25K; a customer-facing AI agent platform that is profitable with 86% gross margin, 60K+ users, and $642K TTM revenue asking $190K; and a mobile AI relationship platform with strong retention, 2,000 active subscribers, and $899K TTM revenue asking $145K. A separate listing is a 12-year-old premium organic skincare brand with 78% gross margin, $387K TTM revenue, asking $348K. — via 1 2 3 4
2. Flippa is listing an AI photo and video app with $29.1K monthly profit, 50% margin, and a 3.3x valuation multiple; it has 278K+ downloads, a 4.6-star rating from 2,800+ reviews, 92% of revenue from subscription renewals, and primarily organic acquisition. Flippa also shared a view from Matt Bradbury that the right time to sell is when a business has shown three consecutive years of strong upward trend with expanding margins. — via 1 2
3. A forwarded post argues founders staring at the same business for years only see bugs, difficult customers, and failed growth attempts, while buyers see customers, brand, product, team, and potential the founder can no longer pursue—talking to buyers can give founders a new perspective on their company's value. Separately, Jordan Calderon grew StratDev 118% year-over-year but decided to hand over the company and start over because he enjoys zero-to-one building more and wants to see if someone more experienced can take it further. — via 1 2

4. Product Launches and Operator Notes

1. Factory launched Teams, which supports adding multiple users to self-serve plans with centralized billing; teams can hand bug fixes, tests, and migrations to Droids so they can focus on downstream development. — via 1
2. Tony Dinh says running a remote software company is great for traveling and working, though he is unsure how long until AI fully takes over software work. He also says his wife's recent social media presence has brought significant brand value, and he is offering to pay for TypingMind branding to be shown in AI-related demos or videos, contactable via DM. A quoted post describes using Claude Opus 5.5 to render New Zealand travel photos in an acrylic painting style entirely in JavaScript, with impressive results. — via 1 2
3. Starter Story profiles Tom, who built a product for tradies reaching $50K monthly revenue, calling tradespeople ideal first customers: pragmatic and unwilling to pay for anything that does not pay for itself, and they came to the team proactively, which shaped the product. Separately, 24-year-old accountant Jake Glacer built an alarm app that requires push-ups to turn off, reaching $50K monthly revenue after four months. Glacer has an accounting background, passed the CPA, worked one year at a Big Four firm, coded on the side for four years through multiple failed startups, and realized his bottleneck was marketing, not development; the product idea came from his own life. — via 1 2
4. Nick Huber argues that almost all US cities remain severely short of "old-school" trades businesses, with weak competition: operators do not invest in marketing, owners resist change, and order books are full and exhausting, so simply being slightly better creates a clear competitive opening. He estimates that if you call five local home improvement companies to build a deck or renovate a kitchen, only one in five will answer the phone, and only one in ten will have time to schedule and show up within three months. — via 1